help with fine tuning

I am trying to finetune GearBindP model to our own dataset but I am struggling to understand how to prepare the data for it.

I am trying to follow https://github.com/DeepGraphLearning/torchdrug/blob/master/doc/source/tutorials/pretrain.rst but is there anything else you suggest?
